Hyperspectral imaging for early identification of strawberry leaves diseases with machine learning and spectral fingerprint features

Anthracnose and gray mold are two most devastating diseases of strawberries which can spread to healthy plants in short time and can cause large-scale yield losses worldwide. However, early identification of anthracnose and gray mold in strawberries is challenging due to that they rise fast and their course is short. Early identification of anthracnose and gray mold in strawberries is of great significance for managing strawberry production, achieving precise target spraying, avoiding the large-scale spread of disease as well as improving the yield and quality of strawberries. In this study, six machine learning-aided methods were developed based on the selected spectral fingerprint features for early identification of anthracnose and gray mold in strawberries using a hyperspectral imaging system. First, infection strawberry leaf dataset was artificially prepared by an expert, and the hyperspectral images (during the spectrum range of 400–1000 nm) of heathy, anthracnose-infected and gray-mold-infected leaves (149 of each type), and fungus-infected leaves respectively had three stages of infection (24 h:43; 48 h:47; 72 h:59). Second, the full spectra of ROI were extracted, and chemometric methods in spectral domain were used to explore the spectral fingerprint features for early identification of anthracnose and gray mold with machine learning. Third, six classification models for identification of anthracnose and gray mold in strawberries were developed, and the classification performance were evaluated and compared. For early detection of anthracnose and gray mold in strawberries, most classification models obtain relatively good accuracy (100%) and robust performance, recognizing the asymptomatic fungus infections classes before the obvious signs of disease appear notably in the strawberry. This study provides a foundational basis for the development of the rapid online inspection as well as the real-time monitoring system for field plant disease detection.